UV 365 vs 395 nm: choosing the wavelength for fluorescence inspection

A fluorescence step lives or dies by the excitation wavelength. Pick 365 nm and most parts glow hard; pick 395 on the wrong material and the image comes back nearly black, or violet haze buries the signal. This guide is for engineers adding UV inspection to a production line: dyed sealants and adhesives, coatings, oils, contamination, optical brighteners. By the end you will know which wavelength excites which materials hardest, what filter cut-on each one needs, why strobing beats continuous drive for heat and LED lifetime, and how to settle 365 vs 395 on your own parts before you order anything.

Why 365 usually excites harder

Fluorescence is a two-step exchange of light. The part absorbs an excitation photon and re-emits a second photon at a longer wavelength, usually in the visible; that shift is the Stokes shift, and it is what lets a filter separate signal from excitation. How much the part absorbs at each wavelength comes from its absorption spectrum, and organic fluorophores - oils, adhesives, resins, brighteners, biological residue - absorb more strongly the shorter the wavelength. That is why the shorter wavelength wins on most materials, and why a part can fluoresce brilliantly under it and show almost nothing under violet.

Photon energy adds a smaller edge:

Each 365 nm photon carries about 8 % more energy than a 395 nm photon (3.40 / 3.14 = 1.08). The absorption coefficient dominates brightness, so treat that 8 % as a tiebreaker, not the reason. Where these bands sit in the wider spectrum is mapped in the wavelength ranges guide, and the fluorescence inspection guide walks the full excitation-and-emission method both wavelengths plug into.

What the filter has to block

The camera must see emission and never excitation. A long-pass filter on the lens does that job: its cut-on sits above the excitation band and below the emission peak. Start at a 400-415 nm cut-on for 365 nm excitation, which leaves a 35-50 nm guard band; start at 420-435 nm for the violet band, leaving a 25-40 nm guard. Two consequences follow.

First, everything below the cut-on is dead to the camera. Excite at 395 and any fluorophore that emits violet-blue between 400 and 430 nm is discarded together with the excitation; that part is a short-wave-only job.

Second, LED output is not a single line. UV LED spectra run roughly 10-20 nm wide and carry longer-wavelength tails, so a short-wave head can leak a little visible violet that a shiny part bounces straight into the lens. If the background will not go dark, add a cleanup bandpass or short-pass on the light itself. A violet head needs only the one lens filter, because its whole output sits inside the blocked band - that single-filter simplicity is the practical case for violet.

If the task is seeing the UV itself rather than the glow, the set-up changes; the reflected UV vs UV fluorescence guide separates the two.

365, 395 and 405 side by side

The spot lights category lists 22 models at 365 nm, 20 at 395 nm and 10 at 405 nm, so all three are real choices, and the light finder narrows the same list by lit area and voltage. The decision values in one view:

Worked example: strobing a 4 W spot light on a 500 ms cycle

The inputs, and where each comes from: the light is a 24 V spot head rated 4 W (its catalog card, shown below); the camera exposure is 2 ms and the line presents a part every 500 ms (example station numbers - substitute your own exposure and takt); the pulse is 3 ms, which is the 2 ms exposure plus 1 ms of margin for trigger delay and LED turn-on. Optical rise times run in microseconds, so that margin is generous; the rise-time guide shows how to measure the real number for your head.

The rule: the light only heats while it is on. Duty cycle is pulse width divided by cycle time, and average power is duty cycle times rated power:

The result: strobing cuts the average thermal load from 4 W to 24 mW, a factor of 4 / 0.024 = about 170. The decision follows from the same arithmetic - size cooling for average power, and treat about 2 W average (a 50 % duty cycle on this 4 W head) as the line above which the station needs continuous-duty thermal design. Matching duty cycle between light and controller is set out in the strobe duty cycle guide.

The head in this example:

For more peak light, the high-luminance line offers a 6 mm high head driven at 5 W:

Drive it from a strobe controller

UV LEDs convert less of their electrical input into light than visible LEDs, so more heat stays in the housing - and heat is what ages them. The same duty-cycle arithmetic that limits the part's dose keeps the junction cool, which is also why strobe operation extends the L70 lifetime; the L70 guide has the mechanism. Run UV heads from a light controller, not a fixed supply, and take the strobe trigger from the camera or PLC; the smart camera strobe wiring guide covers the wiring.

A small spot pairs with a 4-channel, 200 W digital control box that carries a strobe feature:

A station pairing two UV heads needs more headroom:

Put the 365 head on one channel and the 395 head on the other, and the controller sequences them with no extra hardware.

Which wavelength wins for which part

• Dyed sealants, adhesives, oils and greases absorb broadly across UVA and violet: expect 365 to image brighter, with 395 close behind on strong responders.
• Optical brighteners in paper, textiles and some plastics respond hardest at the short end.
• Materials whose absorption only begins near 400 nm do equally under both; pick the violet band for the cooler head and the single-filter set-up.
• Fluorophores emitting violet-blue below roughly 430 nm are a short-wave-only case - a 420 nm cut-on would eat the signal.

An unknown part gets tested, not argued. Mount a 365 nm head and a 395 nm head over one fixture, sequence them through separate controller channels so the camera captures one image per wavelength per cycle - the multi wavelength sequencing guide shows the timing pattern - then compare defect-to-background contrast and keep the wavelength with the wider grey-level gap. Picking the band that carries the signal is the same discipline as the SWIR vs NIR vs visible choice, and it ends the same way: measurement, not preference.

For a confirmed violet-only response, the 405 nm option is a compact spot light with a PNP trigger input:

Camera, lens and filter set-up

Use a monochrome camera. In a narrow band, a colour sensor's Bayer filter discards most photons before they reach a pixel, so the same sensor in mono collects roughly three times the signal. The ambient light rejection acceptance test shows how to qualify the whole chain against room light.

The lens is the easy half. Fluorescence emits in the visible, so a standard C-mount lens transmits and focuses the signal normally; the glass transmission and focus shifts covered in the UV and SWIR lens guide only matter if you image the UV itself. Do fit the lens long-pass filter, and refocus with the filter in place, because its glass thickness shifts the focal plane slightly.

Set exposure on a live continuous image, then keep the strobe pulse at exposure plus margin, exactly as the worked example does. Shroud the station: fluorescence is faint next to room light, and an enclosure beats fighting ambient in software.

Checks before you order

• Image ten known-good and ten known-defect parts under both wavelengths at identical exposure and compute the grey-level gap between defect and background; the test-contrast-before-buying guide gives the protocol.
• If the background stays high, try a higher long-pass cut-on - the emission may start lower than you assumed.
• Redo the duty-cycle arithmetic with your real exposure and takt, and check the controller's per-channel power against the head.
• Match trigger polarity to the head's input (PNP versus NPN) before wiring day.
• Plan the shielding: no direct view of the lit spot, and no reflective fixtures bouncing violet toward eye height.
• Keep dust off the filter and the head window; dust scatters violet into the lens and costs contrast.

When the usual answer is wrong

• The image stays dark even at the short wavelength: push peak brightness with overdrive pulsing, which drives the LED beyond its continuous current for the few milliseconds of exposure; the hidden strobe technique shows how it stays invisible to operators. If overdrive does not rescue it, the fluorophore concentration is the problem, and the fix is on the part, not the light.
• Violet haze survives the filter: that is the visible tail of the short-wave head, and it wants the light-side cleanup filter from the filtering section.
• One spot light sized for a 20 mm field cannot cover a 100 mm field: fix coverage in the finder first - contrast problems are often coverage problems.

The decision

Run the two-wavelength test, then buy in this order. If 395 matches 365 contrast within a small margin, take the violet band: the same image, a cooler head, one filter. If the part only wakes up under the short wave, take 365 with a 400-415 nm long-pass and add the light-side cleanup only if the background demands it. Take 405 only when a violet response is proven. Drive the winner from a strobe controller sized on the duty-cycle arithmetic above.

Next step: narrow the lights in the light finder by wavelength and lit area, size the drive in the controller finder, or start at the lighting step of the configurator. For a quote on a two-wavelength test station, contact us - and the lighting topic page collects the rest of the lighting guides.